Modeling the Effects of Area Closure and Tax Policies: A Spatial Model of the Hawaii Longline Fishery. Ujjayant Chakravorty and Keiichi Nemoto*

Transcription

1 Modeling the Effects of Area Closure and Tax Policies: A Spatial Model of the Hawaii Longline Fishery by Ujjayant Charavorty and Keiichi Nemoto* Abstract We develop an economic model for a multi-species fishery that incorporates the spatial distribution of effort and fish stocs. Catchability coefficients and initial stocs are estimated from catch and effort data for each specific location. Vessels are allocated over space and time to locations of maximum profit which decline with harvest because of stoc externalities. A supply function for labor allocation in the fishery is estimated. The simulated model is applied to the Hawaii longline fishery. The economic impacts of regulatory policies such as reduction of inshore gear conflict and conservation of offshore turtle populations is examined. Key words: spatial-dynamic models, fisheries regulation, area closures, fishing effort, longline fisheries Running Head: Effects of Area Closure and Tax Policies *Respectively, Emory University, Atlanta and University of Hawaii at Manoa. Address for correspondence: U. Charavorty, Department of Economics, Emory University, Atlanta, GA 30322, unc@emory.edu. We owe a major debt to KinPing Tse for research assistance during the early phase of this research. We than Marcia Hamilton and Sam Pooley of the National Marine Fisheries Service Honolulu Laboratory for access to NMFS data and to Chris Boggs, Pierre Kleiber, John Sibert, Mie Travis, John Yanagida and especially Sam Pooley for numerous suggestions and insights. This wor was funded by the National Oceanic and Atmospheric Administration Project No administered through the Joint Institute for Marine and Atmospheric Research, University of Hawaii. 1

2 Modeling the Effects of Area Closure and Tax Policies: A Spatial Model of the Hawaii Longline Fishery 1. Introduction Individual Transferable Quotas (ITQs) have been heralded by many fishery economists as the panacea for the management problems that beset most of the world s fish stocs. In a comprehensive review of the regulatory experience with ITQs, Squires, Kirley and Tisdell (1995) suggest that many countries have preferred input controls to ITQs in fisheries with multiple species and bycatch problems as well as in situations where the costs of monitoring, enforcement and resource assessment are significant. Other second best management measures such as area closures and gear restrictions have been adopted in more complex fisheries in several countries such as the United Kingdom, Norway and Italy. Area and seasonal closures have also been popular in the management of migratory species such as tuna in the Atlantic and Pacific Oceans (Gribble and Dredge, 1994). There is a large body of literature (see survey by Townsend, 1990) that examines the economic and biological impacts of alternative regulatory policies such as ITQs and various forms of limited entry programs (e.g., gear restrictions). Most published studies have either focused on a single regulatory instrument such as a quota on harvest or gear restriction or a combination (Moussali and Hilborn, 1986; Stollery, 1984). Other analyses have attempted to endogenize the length and timing of closures in a programming model of stoc recruitment (Watson, Die and Restrepo, 1993). These models, by and large, implicitly assume away substitution effects of area closures, i.e., when a certain fishing ground is closed, effort may be reallocated elsewhere. Several studies suggest that these substitution effects may be quite significant. For example, Cadrin et al. (1995) point out that when stocs are migratory, closure of inshore areas to more efficient vessels may not confer the expected benefits because fishing effort will be displaced to unprotected areas. Closure of inshore fishing areas led to increased allocation of effort to onshore fishing grounds in the Gulf of Mexico brown shrimp fishery (Gracia, 1997). Very recently, lawyers for the Earthjustice Legal Defense Fund concerned about turtle bycatches in the North Pacific successfully argued that substitution of boats to other regions will minimize the impact of a 2

3 moratorium on fish harvests in the North Hawaiian Ocean. As a result, U.S. District Judge David Ezra too the unprecedented step of restricting Hawaii longliners from fishing north of 28 degrees latitude. (TenBruggencate, 1999). In what follows, we examine the economic impacts of area closure and tax policies by developing a model which explicitly incorporates both the spatial and dynamic elements. The model has several unique features that include (i) estimation of the catch-abundance relationship for a multispecies fishery by spatial location (ii) sequential allocation of vessels over a spatial grid using a crew-profit maximizing criterion, and (iii) estimation of a labor supply function based on the labor-leisure trade-off. The model is applied to the spatial and dynamic allocation of longline vessels in the Hawaii pelagic fishery. Model results are found to predict actual vessel allocation data reasonably accurately. The model is then used to generate economic impacts under alternative area closure restrictions and tax policies. The differential impact of area closure policies that reduce inshore gear conflict and turtle bycatch in offshore fisheries is compared with an increase in the tax on harvest. It is found that harvest taxes have a minimal impact in achieving conservation objectives relative to area closures. Harvest taxes have little effect on industry revenue but reduce crew wages and boat owner incomes by about 20 percent. Area closure policies that reduce turtle by-catch in the northern latitudes also bloc access to lucrative swordfish fishing grounds, but they have a relatively small impact on vessel profits. This is because boats are able to switch to inshore fishing areas. Policies that reduce gear conflict in inshore areas have the smallest impact on crew income and result in reduced harvests of the major inshore species. While the individual elements of the proposed model are not novel, we believe that the inclusion of a spatially non-uniform distribution of fish stocs and harvesting costs in obtaining the equilibrium allocation of vessels over time and space helps develop a modeling framewor which is powerful in predictive capacity and policy analysis. The spatial feature allows for differential accounting of vessel travel costs from port and fishing costs by location. The dynamic nature of the model allows for stoc externalities and exogenous changes in demand and cost parameters 3

4 and in the longer term, discounting can be easily incorporated. In future wor, biological information on the spatial migration of pelagic fish stocs can be included to determine instantaneous fish stocs net of harvest and migration. Extension to multiple ports of origin and political jurisdictions is straightforward. The paper is organized as follows: Section 2 details the elements of the proposed model. Section 3 describes the calibration of the model using data from the Hawaii pelagic fishery. Section 4 demonstrates model application by examining impacts of regulatory policy changes. Section 5 concludes the paper. 2. The Model The model is based on a standard framewor of maximization of fleet profit in the short run, where the allocation of fishing effort is determined over space based on the comparison of net revenues from each fishing location. The fishery can be thought of as a regulated open access fishery, where seasonal closures are used to achieve regulatory objectives such as species and bycatch conservation. The net revenues in turn are dependent on stoc sizes of each species and allocated fishing effort in fishing locations as well as exogenous fish price and catchability of each species. Price and various stoc sizes across fishing regions in each period influence the fleet revenue per trip, while the distance between the harbor and each fishing area affect travel costs. As more trips are allocated into a particular fishing area, the expected revenue per trip from the area diminishes due to the stoc externality. Boats are divided according to whether they target swordfish or tuna, as explained below, which in turn affects the catchability of different species. In equilibrium, the average net returns per trip are equalized across locations. Catch Function Consider a fishery in which there are I species of fish (e.g., swordfish, bigeye, and yellowfin) denoted by i = 1,...,I and K fishing areas or locations, indexed by = 1,...,K. Stocs of different species are assumed to be nown at the beginning of a given time period in each area. These time periods are chosen to be sufficiently short such that in each area, stoc changes within a period 4

5 due to fish reproduction, natural mortality, and stoc migration across areas can be ignored; i.e., fish stocs within a period are expected to decrease only due to harvesting. Let the catch function over this unit time period be given by f ( E, γ i ) Ci, = ( 1 e ) B i, (1) where C, is the catch of species i in area, g i is a vector of the catchability coefficient for i species i, E is a vector of fishing effort (e.g., number of hoos) in area, and B, is the fish i stoc for species i in area. In (1), the instantaneous fishing mortality rate, f, is defined as a linear sum of effort by set types, ( γ ) f E, = γ, E, s, (2) i i s s where s represents alternative set types (e.g., tuna or swordfish set), γ i,s and E,s are respectively, the catchability coefficient and the amount of fishing effort by set type s. Vessel trips need to be differentiated by whether they target swordfish or tuna (or both), since the cost structure and the catchability vary depending on the species targeted during the trip. A slightly more generalized form of (1) has been used by Clar (1985) and Deacon (1989) ( ) although they do not consider heterogeneity in the targeting of species. The term ( 1 e ) f E,γ i is defined as the fishing mortality rate that depends on the level of labor, the particular species targeted, the type of boat and gear used. Its magnitude is always less than unity. Several restrictions are imposed on the above catch function. We assume that the fish stoc is infinitely diffusive within area ; i.e., the density of fish is linearly related to the residual stoc size and is distributed uniformly within each fishing area (Clar, 1985). 5

6 Revenue from a Fishing Trip We assume that ex-vessel fish prices are given. Although not attempted here, a demand function for each species can be substituted at the cost of additional model complexity. Implicitly, we assume that each boat has perfect nowledge of the fish stoc in area. Then the total revenue from fishing in area is obtained by summing over revenues from each species. The average revenue per trip in area, AR,, can then be expressed as Trip AR Trip, Pi Ci, = i N = i f ( E, γ i ) Pi ( 1 e ) B N i, (3) where P is the price of species i and i N denotes the total number of fishing trips to area per unit time period. It is easy to see from (3) that the average revenue will increase with fish prices and initial stoc, and decrease with the number of trips. We assume that all boats that target the same species (tuna or swordfish) are identical and that once the boat reaches a fishing location, they expend the same fishing effort (e.g., number of hoos per day over an equal number of days). The assumption of identical effort across boats targeting tuna or swordfish may be reasonable for an industry with homogeneous gear (e.g., longliners) in which most boats would have roughly similar storage capacity for bait and harvested fish. Second, the time spent fishing may be limited since the quality of the freshly harvested fish on board begins to decline rapidly with time. Let FD denote the number of fishing days per trip. Since each boat can change types (e.g., targeting swordfish or tuna) during a single trip, FD =,, where FD,s is the number of s FD s days a boat is of type s per trip in area. It implies that the numbers of days a boat chooses to target tuna or swordfish may vary but the total number of fishing days per trip is fixed. The allocation of set type within a trip is a choice variable. Let E s denote the amount of fishing effort 6

7 (sets of 1,000 hoos per day) by a boat of type s, which is constant. Then the aggregate amount of fishing effort in area by set type s, is E E FD N. (4), s s, s Cost Structure and Crew Wage We assume ris neutrality on the part of boat owners and crew, but incorporate features of the labor-employment relation that have a bearing on apportioning of revenue and costs from fishing (Plourde and Smith, 1989). For this purpose, we classify the costs of fishing into two categories: fixed and variable (i.e., operational) costs. In the share system that is prevalent in the Hawaii longline fishery (Hamilton, Curtis, and Travis, 1996), fixed costs are usually borne by the vessel owner and consist of overhead expenses (e.g., maintenance, mooring, and depreciation charges) which do not depend directly on the fishing trip and are more or less fixed on an annual basis. Therefore, we assume that fixed costs do not affect the trip allocation decision in our short run model. The variable costs are the expenses incurred during the fishing trip (e.g., fuel, bait and gear). This can be further broen down into costs incurred while traveling to area and fishing in that location. We assume away travel costs within area, which may be reasonable because they are liely to be a small fraction of the costs of traveling from port. For simplicity, each trip involves direct travel to the chosen destination and return to port, i.e., fishing at multiple locations within the same trip is not allowed. In addition there is an auction fee that is levied as a percentage of the total catch. These variable costs are shared equally by the owner and crew. The owner is usually absentee while by crew we denote all hands on board including the captain. The net revenue from a trip to area, NR is NR = ARTrip, ( 1 τ ) as FD, s + b TD (5) s 7

8 where cost parameters a s and b represent the average daily variable costs of fishing with set type s and traveling, respectively and TD is the number of days spent in traveling from port to destination and bac. As mentioned before, since all boats spend an equal number of fishing days per trip, the total number of fishing days per trip FD is equal across trips. On the other hand, the number of travel days by the average vessel speed. TD can be estimated by dividing twice the distance to area from port The incomes accruing to the vessel owner and crew from a trip to area are given by OI ( ) = 1 λ NR (6) and CI = λ NR (7) where OI and CI and ( λ ) 1 and λ denote the respective incomes and relative shares of net revenue accruing to the owner and crew. The crew income from a trip in (7) or the crew wage per day, which is the crew income divided by the total number of trip days, may be used as a measure of wage. However, both indicators may be biased when comparing wages across fishing areas due to the variation in the distance of the fishing location from port. We remove this bias by normalizing wages with respect to distance employing a procedure detailed in Appendix 1 which yields NCW, the normalized crew wage for a trip to area. Finally, let NCW denote the weighted average normalized crew wage for the entire fishery per unit time period, obtained as NCW = NCW N N. (8) Finally, labor allocation per period in the fishery, measured in vessel-days, is obtained by summing both fishing and traveling days over all trips to all fishing areas as 8

9 ( ) VD = FD + TD N (9) where VD is the aggregate amount of labor allocation in vessel-days per time period. Given exogenous fish prices, a derived demand function for labor is obtained through optimal allocation of vessel trips over space. Since vessels are assumed identical in every respect, it is straightforward to distribute the representative vessel trip spatially starting from the fishing location that yields the maximum normalized crew wage to the crew (i.e., NCW ). Note that the allocation of sets (tuna and swordfish) within each trip is endogenous. As vessels get assigned to locations generating maximum normalized crew wage, subsequent harvests in that location decline because of stoc externality. Other competing fishing locations become more profitable. Thus as the aggregate number of trips - all of which consist of an equal number of fishing (though not travel) days - increases, the marginal, and hence the average trip wage decreases. Because of the discrete nature of the problem of allocating a fixed number of vessel trips, the equilibrium normalized crew wage will only be approximately equal across locations. Thus in a model where the number of boats is infinitely divisible, NCW and NCW are exactly equal. A plot of the normalized wage index as a function of aggregate industry effort (vessel-days) yields a downward sloping derived demand function for labor. Note that this demand function may shift in response to changes in fish prices or the initial distribution of the fish stoc. Labor Supply and Equilibrium Allocation of Trips over Space The fishing industry is ideally suited for modeling the labor supply behavior of fishermen. This is because most fishermen are self-employed and boat captains generally have the power to decide if and when to undertae a fishing trip and for how long. Furthermore, as pointed out by Gautam, Strand and Kirley (1996), unlie other professions in which labor and leisure activities coexist almost on a daily basis, commercial fishing in the high seas is demanding wor coupled with limited leisure opportunities. This maes the disutility of spending another day fishing or staying at sea an important determinant of the labor supply decision. Assuming a fixed number of fishing vessels, the aggregate labor supply function can be expressed as 9

10 ( ) VD = S NCW, M (11) L where M is income from non-fishing activities. Labor supplied by the crew (including captain) is expected to increase with the normalized crew wage NCW and it may in the short run exhibit bacward-bending properties (as in Gautam, Strand and Kirley, 1996). In general, non-fishery income M may affect the labor-leisure substitution. However, for commercial fisheries as considered here, M may be relatively small, and in any case it is difficult to obtain data for nonfishery income earned by fishermen. It is therefore ignored in the empirical estimation. In order to econometrically estimate the labor supply function in (11), we obtain NCW and VD using actual data (Kennedy, 1992). That is, the actual numbers of catches and vessel trips to area are used in (3) as compared to the use of equations (1-10) in the calculation of the derived demand function obtained from stoc and price data. Note that the derived demand function for labor is expected to shift in response to changes in fish prices or the distribution of fish stocs, while the labor supply function remains constant across periods. Finally, the equilibrium levels of NCW and VD are obtained by equating the derived labor demand with the estimated labor supply function from (11). 3. Application to the Hawaii Longline Fishery In this section we apply the above model to the Hawaii longline pelagic fishery. The Hawaii commercial fishery has rapidly grown since 1987 with annual commercial value estimated at roughly $60 million (WPRFMC, 1997). Longline vessels account for more than 80 percent of gross revenue, the remaining being from baitboat (pole-and-line sipjac), handline and trolling. For the purposes of this study, we only deal with longline vessels which are reasonably homogenous in terms of fishing technology and other vessel characteristics. The major species targeted by the longline fishery are broadbill swordfish and bigeye tuna (He, Bigelow, and Boggs, 1997) while yellowfin and albacore tunas and striped marlin also represent a significant share of 10

11 the total ex-vessel revenue. Together these five species of fish, accounted for approximately 90% of gross revenue in We construct a spatial grid that divides the fishery into 56 five-by-five degree (latitude and longitude) squares centered in the Main Hawaiian Islands (MHI). These areas are located between latitude 5 N to 45 N and longitude 140 W to 170 E. Each five-degree square is defined using its southeast corner as the reference point, e.g., the square between latitudes 15 and 20 N and longitudes 140 and 145 W is labeled as 15N140W (as per a classification system developed by Curran, Boggs, and He, 1996). Catch and fishing effort data from the 1995 longline logboo, collected by the National Marine Fisheries Service (NMFS) (Dollar and Yoshimoto, 1991) are aggregated by five-by-five degree square and by month. The data suggests that 1,125 trips (corresponding to 11,129 sets and about 13.3 million hoos) were taen by 110 active longline vessels in 1995 (WPRFMC, 1997). That is, on average each longline fishing trip lays approximately 10 sets one set each fishing day. Swordfish Vs Tuna Sets One complication that needs to be considered in empirical wor is the targeting strategy of the vessel. That is each boat can target swordfish on any given fishing day, or it may target tuna (bigeye and yellowfin) while the remaining two species albacore and striped marlin are caught mainly as by-catch. Targeting these distinct species imposes distinct fishing and cost characteristics on the boat. For example, swordfish sets are soaed overnight and they use more expensive bait (squid) and other devices (lightstics). Tuna sets are soaed in daylight, and use cheaper samma (saury, Cololabis saira) as bait. The catchability coefficients (from (1)) would be different for each set since swordfish (tuna) targeting would lead to bigger swordfish (tuna) catches, ceteris paribus. Allocation of sets is done by inspecting logboo data and designating each set as swordfish or tuna set depending on whether they satisfy the above qualitative criteria. About 11 percent of the sets did not fall clearly into any category e.g., a set that soaed in the 11

12 night, used squid as bait but did not use any lightstics. These were allocated by looing at which of the above criteria they matched more closely. Given the two fishing strategies or set types, the specification of the instantaneous fishing mortality rate in (2) can be simplified as ( ) f E,γ = γ E + γ E i i1 1 i 2 2 (12) where subscripts 1 and 2 denote tuna and swordfish sets respectively. Since the average number of hoos used in a swordfish set (820) was different from a tuna set (1,498), the fishing effort levels in area by type 1 and 2 can be expressed from (4) as E = FD N and.. Here ( FD 1 N ) and ( FD N ) E = FD N represent the total numbers of tuna and swordfish sets conducted in area and FD 1 + FD 2 = 10 since the sum of swordfish and tuna sets per trip must equal the total fishing days, and we define one unit of fishing effort as 1,000 hoos. Estimating Catchability Coefficients and Fish Stocs Given the two distinct fishing strategies, two sets of catchability coefficients for each species (i.e., for tuna and swordfish sets) are estimated from monthly catch and effort data, details for which are given in Appendix 2. The results are summarized in Table 1. The catchability for swordfish with a tuna set was very low (0.0002) which implies that a tuna set catches very few swordfish. Also for a given stoc size and unit effort, a swordfish set catches 22 times more swordfish than a tuna set. The table also suggests that a swordfish set catches more bigeye and yellowfin tuna than even a tuna set, although fishermen may actually prefer to catch these tunas using the cheaper cost of a tuna set. Oamoto (1999) obtained similar results, where catches of albacore and striped marlin were higher from tuna sets than from swordfish sets, while swordfish catches from swordfish sets were higher than those from tuna sets. 12

13 The catchability coefficients for the five major species as well as monthly effort and catch data are used to estimate the initial size of the monthly fish stoc B i, in each location from (1). That is, the estimated fish stoc (population of species i in location ) is assumed to be in place at the beginning of each month but depletes with harvesting as trips are allocated at each location. A new stoc is estimated at the beginning of each month. In one sense, the model simplifies the inflow and outflow of fish migration in each grid by assuming that migration could only occur instantaneously at the point of transition between successive time periods, i.e., at the beginning of each calendar month. Notice that a calendar month is only an arbitrary device, and the model could be built with weely price, catch and effort data, if available, although allocation of trips which usually last longer than a wee, may be problematic in that situation. Expected Revenue from a Trip Monthly fish prices are assumed to be exogenously determined since most of the fish is sold in marets in Japan and in the U.S. mainland (WPRFMC, 1995). Prices were computed using revenue data for each species collected by the Hawaii Department of Agriculture and Resources (HDAR). Catch data is in terms of numbers of fish caught and fish prices are in dollars per standard-sized fish, as shown in Figure 1. Using (3), expected revenue in each location was computed from fish prices and estimated stocs. Since the aggregate revenue from the five major species was percent of the total revenue reported in 1995, a correction factor of was applied to account for other minor species and side catches. Variable Costs of Fishing As mentioned earlier, we ignore fixed costs of fishing and focus only on variable costs and taxes. Both the auction fee and the excise tax are shared equally between the owner and crew and are a fixed proportion of gross trip revenue. In Hawaii, longliners are legally obliged to sell their catch to the United Fishing Agency Ltd. which charges each vessel an auction fee equal to 10% of the total revenue from a fishing trip. The excise tax rate is an additional 0.5% of the total revenue (Hamilton, Curtis, and Travis, 1996). 13

14 Variable costs, including the cost of food, oil, fuel, bait, light-stic, ice, and miscellaneous gear, can be broen down according to their relationship to fishing or traveling activity. Food, oil, and fuel are consumed for both, while expenses on bait, light-stics, ice, and miscellaneous fishing gear occur only during fishing. Table 2 details the breadown of the variable cost. Based on the survey by Hamilton, Curtis, and Travis (1996), the average daily costs for food and oil are $81.54 and $9.41 respectively. The cost of fuel is on average higher on a travel day than on a fishing day; i.e., the average fuel cost was $ for fishing days and $ for travel days. Another complication is the calculation of variable costs is the higher expense of targeting swordfish relative to the tuna species (i.e., bigeye and yellowfin). The average variable cost per fishing day (out of the 10 fishing days in a trip) depends on the proportion of swordfish-targeted sets in the standardized 10 sets of fishing effort. As indicated in Table 2, the cost of a fishing day is much higher than the cost of a travel day, and swordfish fishing is almost thrice as expensive as tuna fishing. This yields the following formula for variable costs for a trip to area : ( ) VC = r TD (13) sw where r sw is the share of swordfish sets in the total. The above equation implies that the respective daily costs for a tuna and swordfish set are about $780 and $2063, due to more expensive bait, lightstics and miscellaneous gear in the latter. Net Return and Normalized Crew Wage The net return for a trip to area, and the normalized crew wage can then be expressed as NR = i ( γ i1 E 1 + γ i 2 E 2 ) Pi ( 1 e ) B N i, ( ) VC (14) and NCW = 050. NR TD (15) 14

15 where is the total share of revenue paid out as auction fee (10%) and excise tax (0.5%), is the ratio of wages from a travel day relative to a fishing day (see Appendix 1), and the vessel owner and crew each collect 50% of the residual profit (i.e., λ = 0.50) more than 95% of Hawaii-based longline vessels follow the equal sharing rule (Hamilton, Curtis, and Travis, 1996). These equations yield the derived demand for labor as explained in the previous section. Estimation of Labor Supply To get the supply function, the monthly total revenue from trips to area was estimated using actual catch data by region from the 1995 longline logboo. Unfortunately, logboo data does not reveal the number of trips made by vessels to a specific location, only the total numbers of sets and hoos placed in each square by month. Since we assume that the exact 10 sets must be conducted in a single fishing location within a trip, dividing the total number of sets placed in each square by 10 yields the actual number of trips. The number of vessel-days, computed from (10), were aggregated by month, and are summarized in Table 3. The annual number of trips approximated thus was 1,165, somewhat larger than the 1,125 total longline trips made in 1995 (WPRFMC, 1997). One reason for this overestimation was that vessels tended to conduct more sets during distant-water fishing trips. Part of the error could also be due to errors in recording of logboo data. To derive the supply function, a simplified specification of a labor supply function used by Battalio, Green and Kagel (1981) was estimated as follows: ( ) ( ) 2 ln VD = ln NCW ln NCW, (16) (45.03) (21.89) where numbers inside the parentheses are t-ratios. The resulting adjusted R 2 and Durbin-Watson statistic were and respectively. Both parameters were significant at the 1% level. Since the crew would not go fishing if the expected wage is not high enough, lnvd 0 when ln NCW = 0, the intercept in (16) should be zero or negative. Because a positive but insignificant intercept was obtained in the preliminary estimation, the intercept was restricted to zero. The 15

16 labor supply function obtained is illustrated in Figure 2. Supply is increasing with NCW when NCW < $727/day, but it is bacward-bending at higher values. Ex-vessel revenues from the five major species shows that the longline industry earned more than $4 million per month during the first quarter and in December, and much less during the other months, particularly from July to November. The above data indicate that the bacward-bending supply curve may be reasonable. Simulating the Hawaii Longline Fishery Equilibrium monthly effort allocation that equates demand and supply of effort for the entire fishery is determined using a simulation algorithm written in the programming language Turbo C++. Note that there are 12 different demand functions since stocs and prices vary by month but only a single estimated supply function. Results are only shown for March and August 1995, when the derived demands for labor are respectively large and small, are shown in Figure 2. Model results for the baseline year 1995 are compared with the actual distribution of effort in the Hawaii longline fishery, shown in Figure 3. In order to investigate the importance of the leisurelabor tradeoff in the estimation of the supply function, an alternative model in which there is no income effect on the consumption of leisure (i.e., days on shore) is also presented in the figure. The later model ignores the fisherman s disincentive to supply effort brought on by higher trip wages. Without this effect, the supply curve is horizontal; that is, fishing trips are allocated until labor demand is equal to the normalized crew wage computed from 1993 wage data (Hamilton, Curtis, and Travis, 1996). As shown in Figure 3, the model without the labor-leisure tradeoff demonstrated poor fitness. For example, the number of trips allocated was overestimated in the first quarter (particular in March) when trip revenue is relatively high (Table 3) due to higher fish prices and stoc abundance, and underestimated during April to November (no trips were allocated during August and September) when trip revenue was relatively low. This suggests that overestimation may have been caused because the disincentive to supply effort when crew wages were high in the first quarter was ignored. This comparison suggests that incorporation of the 16

17 leisure-labor tradeoff may be significant in modeling fishermen s behavior as suggested by Gautam, Strand and Kirley (1996). On the other hand, the model with the labor-leisure tradeoff performs maredly better in tracing the actual allocation of boats. The Mean Absolute Percentage Error (MAPE) was 11.4%, while that for the case without the labor-leisure tradeoff was more than 70%. However, gaps between the actual and predicted number of trips still remain, as shown in Figure 3. The simulated number of trips is somewhat underestimated from March to May, and overestimated from September to November. A large part of the difference can be explained by the difference in the number of swordfish sets from optimization and the actual spatial distribution as shown in Table 4. For instance, in March and April, only 7 trips are allocated to two fishing regions near the Main Hawaiian Islands (MHI: 20N155W and 15N155W) although more than 40 trips were allocated in reality. However, relatively more trips are allocated in other areas, such as north of 30 N. On the other hand, from September to November, 47 more trips are allocated to the two MHI regions in the simulation while 30 less trips are allocated to the regions north of 30 N. Part of the error may be due to the fact that vessels tend to fish in familiar locations, not necessarily in those which return maximum profits. There is also the difficulty of estimating catches for each species, as seen in the species-wise breadown of actual and simulated catches given in Table 5. It shows that the simulated aggregate catch of albacore and striped marlin were higher by more than 20%, although the error margins for the other three major species (i.e., swordfish, and bigeye and yellowfin tunas) was less than 10%. One possibility is that fishermen may not be explicitly incorporating revenues from albacore and striped marlin in their decisionmaing process because these two are somewhat undesirable species, much less valuable than the other three and tae up scarce storage space in vessels (Kelleher, 1997). Some other factors that may contribute to the difference in results are (i) the assumption of identical vessels in terms of cost structure, speed, and other parameters such as the number of 17

18 light-stics used per fishing set, (ii) perfect nowledge about fish prices and stocs in each fishing location, (iii) the assumption of costless switching between tuna and swordfish sets, and (iv) restriction to one fishing location per trip. In particular, certain group of vessels use only one strategy (tuna or swordfish) for a long period of time due to factors such as personal preference and other vessel-related physical constraints. 4. Policy Simulation We use the model to examine the economic impacts of three proposed regulatory policies: (i) the closure of two five-degree squares, including the fishing areas off the main Hawaiian islands to avoid gear conflicts (ii) closure of all fishing areas north of 30 N for sea turtle conservation, and finally (iii) increase of auction fee from 10% to 20% for revenue generation. Reducing Gear Conflict: Closure of Areas near the Main Hawaiian Islands (Case 1) Limited entry restrictions such as area closure, are particularly appropriate in reducing short-run (or crowding) externalities (Townsend, 1990). Several gear types often compete for the same species of pelagic fish and hence the exclusion of a particular gear type would reduce crowding. In Hawaii, longline and surface fleets (trollers and handliners) have often fished in the same locations, especially within 20 nautical miles of the shore (Sillman, Boggs, and Pooley, 1993). Historically, longline vessels have been excluded from fishing in certain regions from time to time. In recent years, troll and handline landings of several pelagic species (e.g., yellowfin) have declined substantially, while longline landings have increased (Pooley, 1994). Other small commercial, charter, subsistence and recreational boats operating near-shore have also been adversely affected. We examine the impact of the year-round closure of two areas, 20N155W (including Oahu and Maui) and 15N155W (including a major part of the fishing areas close to the Big Island). The results are summarized in Table 6. The importance of the closed areas can be seen from column A in the Base Case for example, in August, October and November, more than 50% of the effort was centered on the two closed squares. The column (C - B) in Table 6 shows the substitution of vessels into other areas as a result of area closure. Vessels compensate by fishing in distant 18

19 waters, which results in an increase in costs and reduced wages as summarized in Table 9. The aggregate number of trips declined. Originally 322 trips were made into the closed areas, but with the closure policy, only 180 were reallocated, leading to a net decrease in 142 trips out of the original 1192 trips - a decline of about 12 percent. Conserving Sea Turtles: Closure of the North Fishing Areas north of latitude 30 N (Case 2) Interaction between longline gear and endangered species such as sea turtles is continuously reported in the logboo data (Ito, 1995). Leatherbac, green sea, loggerhead, and olive ridley turtles were reported to have been accidentally caught a total of 84 times during 1994, although these interactions are widely believed to be under-reported. Kleiber (1998) estimates that approximately 700 sea turtles were taen and around 100 were illed in Most of the loggerhead and leatherbac turtles were caught in the areas north of 30 N. Fishery biologists such as Nitta and Henderson (1993) have suggested the closure of fishing areas north of 30 N to conserve sea turtles. The impacts of such a policy are simulated in Table 7. Unlie the previous case, turtle conserving policies confine longline vessels to fishing areas closer to Hawaii, i.e., below 30 N. Since more vessels now fish inshore, travel days decrease, which in turn enables more trips to be taen. Thus, aggregate number of trips shows a small increase. The ratio of swordfish sets to the total falls from 31% to 20%, and hence swordfish catches, which mostly occur in the high seas, decline significantly (40 %) causing normalized crew wages to fall by 9.4 % (Table 9). Increasing auction fee from 10% to 20% (Case 3) Increase in the auction fee may serve as a mechanism not only to reduce the profitability of fishing and thereby preserve fish stocs but also as a means of generating additional revenue for the State. We thus examine the impacts of an increase in the auction fee rate from 10% to 20 % (Table 8). Since the auction fee is a fixed percentage of total revenue before netting variable costs, increase in the fee results in a disproportionate negative effect on boats specializing in distant-water fishing whose variable costs tend to be higher. In particular, trips fishing north of 30 N are significantly affected (down by 31 trips), followed by those fishing between 5 N and 30 N, as shown in Table 8. 19

20 Counter-intuitively, because of the bacward bending nature of the supply function, a higher auction fee increases the total effort level and the total number of trips when wages are higher especially in the first quarter and in December. Correspondingly, effort reduction will be higher with auction fee when labor supply is positively sloped, i.e., at low wage levels in the summer and fall (see Table 8). Comparison of the Three Policies The differential impacts of the three policies on trip allocation, employment, total revenue, government income through taxation, shared costs, profits and wages are summarized in Table 9. All three policies reduce industry revenue. The total revenue from all longline trips will decline under all three policies although the auction fee causes owner and crew incomes to decline significantly (21%). Annual income accruing to vessel owners is simply 50 percent of the fleet profit divided by the number of active vessels. Thus boat-owner incomes are most seriously affected by auction fees. Sustained low income may cause the net profit to fall below industry long-run average fixed cost, leading to exit from the industry. Some instances of exit from the Hawaii longline fishery to the U.S. Gulf Coast and Fiji have been observed in recent years (Travis, 1998). Crew incomes per trip are more negatively affected by turtle conservation (11 %) and least by reducing gear conflict. However, this comparison is overstated because the average trip length is shorter under turtle conservation (boats fish closer to shore) than under gear conflict regulation. On the other hand, crew wages per day are more affec-ted by gear conflict regulation (12%) than by turtle conservation. However, normalized crew wage, which is net of travel days, is least affected by gear conflict policies (7.2%), suggesting that turtle conservation has more of a negative impact on the crew than gear conflict policy. The effect of alternative policies on conservation of fish stocs is also shown in the table. Interestingly, reduction of gear conflict leads to a significant reduction in catches for all the four species other than swordfish since they dominate harvests close to port. However, turtle conservation leads to a significant reduction of swordfish catches, but not in others. Rather, 20

21 catches of bigeye and yellowfin tunas and striped marlin actually increase. Although an auction fee hie will mean more tax revenues, the impact on fish conservation is minimal since there is very little substitution of vessels across locations and the increased fees only helps in reducing net revenues. 5. Concluding Remars This paper develops a spatial and dynamic model of the allocation of fishing effort that explicitly incorporates the spatial distribution of multiple fish stocs, stoc externalities from fishing and the relationship between fishing effort and crew wages normalized across fishing locations. The model is used to simulate the monthly allocation of effort in the Hawaii longline fishery for the year It is shown that estimation of a labor supply function significantly improves model prediction. The impact of regulatory decisions such as inshore (reduction of gear conflict) and offshore (reduction of turtle bycatch) area closures and taxes on harvest are examined. Inshore area closure leads to vessels moving to more distant waters. While vessel-owner s incomes decline significantly, the effect on crew income per trip is smaller since they benefit from taing longer trips. Catches of inshore species is significantly reduced. Offshore area closure policies reduce swordfish catches by about 40%. Average trip lengths decline by about 20% since boats fish closer to shore. On the positive side, substitution by boats into inshore areas and an increase in the number of trips allows for a relatively small adverse impact on fleet profits (5.5%). An increase in the auction fee on harvest succeeds in simming profits from the fishery and nearly doubles tax revenues but has little effect on conservation of fish stocs. These results could be useful in assessing the impacts of regulatory policy on industry groups or on conservation objectives (Wilen, 1993). There is an interesting asymmetry between the two area closure policies: reduction of gear conflict and turtle conservation. In the first, near-shore area closures lead to a smaller amount of trips but of longer duration. The crew continues to receive wages from the increased travel days, but that does not benefit the boat owners. Thus inshore area closures have a bigger impact on incomes accruing to boat owners. On the other hand, distant area closures such as turtle conservation policies lead to a larger number of lower duration trips 21

22 and has the opposite effect. Fiscal policy instruments such as auction fee increases however, have a significant effect on both parties because there is limited scope for substitution, while trips yielding marginal returns are no longer profitable. The other major impact of turtle conservation is the increased harvesting of near shore species such as yellowfin, bigeye and striped marlin. Although not considered in the model, this can adversely affect catches by competing fleets such as handliners and trollers as well as recreational vessels. However, nearshore area closures increase swordfish catches marginally but have a positive impact on the stocs of the competing species. Our results can be used to compute the rough implicit price of saving a loggerhead turtle. For example, Kleiber (1998) estimates that 66 loggerhead turtles were illed through interaction with longline gears in Then using our model results, the rough cost of adopting turtle conserving policies in terms of foregone profits to the longline fleet is approximately $14,924 per turtle. These types of implicit valuations can be used by policy maers to analyze tradeoffs and mae appropriate policy decisions. The model developed in this paper can be improved in several different ways. The initial fish stoc size is estimated based on current catch and hoo data, assuming no inflow or outflow within a period but allowing for stoc changes across each period, i.e., between successive months. Possible extensions include incorporating a migration function that allows for locational stoc movements that are a function of stoc differentials between adjacent grids. This function may display seasonal variations based on biological information on pelagic fish movements. Possible improvements in future research include estimation of the labor supply function using several years data. Additional factors affecting fishing effort supply (e.g., time lag, income from non-fishing activities) could also be modeled in later wor. The model does not account for interactions with other gear types (e.g., handline and troll). For example, moratorium on distant shore fishing will increase inshore fishing by longline boats and may increase incidence of gear conflict. Lastly, the price and catch data used is for the most recent year (1995) available while 22

23 the cost data is for This asymmetry could introduce errors in simulation although it is implausible that the cost structure may have changed significantly within two years. New costearnings data from recent NMFS surveys could be used to further improve the predictive power of the model. 23

24 Appendix 1 Calculation of Normalized Crew Wage Dividing crew income by trip length will yield biased estimates of crew wages since both fishing and travel days are included in computation of trip length and trips to different locations will entail different travel times. Given the nature of activities on board, wages per fishing day are expected to be higher than wages per travel day. To obtain this relationship, 95 available observations from 1993 trip data collected by Hamilton, Curtis, and Travis (1996) was used to regress the average crew income per trip with the average numbers of fishing days and travel days per trip as AW = FD TD (A1) j j j (9.21) (2.80) where AW j, FD j and TD j are the average crew income and numbers of fishing days and travel days per trip for vessel j, respectively. The above t-ratios suggest that both parameters were statistically significant at the 5% level. Although the number of fishing days per trip is fixed in our model, there was some variation on average number of fishing days in the data. The mean and standard deviation were 10.6 and 2.9, respectively. Since the Breusch-Pagan test rejected the null hypothesis of homoscedasticity, heteroscedasticity was corrected using Shazam econometric computer software (White, 1993). The intercept was negative and insignificant in the preliminary estimation, hence was restricted to zero. The above results imply that the expected wage per travel day is 38.04% (= /917.17) of the wage from a fishing day. Therefore, we remove this bias by calculating the normalized crew wage (NCW) for a trip to area as the crew income at location divided by the effective trip length as follows: NCW = CI. (A2) FD TD 24

25 Appendix 2 Estimation of the Ratio of Catchability Coefficients From (12) define ( ) X ( ) X f E,γ i = γ i1 E1 + γ i2 E 2. Then the catch function (1) can be expressed as C = 1 e B. Since X is usually between zero and unity (Deacon 1989), we can now i, i, expand ( 1 e X ) around X = 0 as a Taylor series to get X X X X X 1 e = + + 1! 2! 3! 4! Λ (A3) which yields a modified catch function ( ) C = B α X = B α γ E + γ E i, i, i, i, i, i1 1 i2 2 (A4) in which α 2 3 X X X ( X ) = ! 3! 4! Λ. (A5) where α is positive and monotonically decreasing with X from a maximum value of unity. (A5) implies that catch-per-unit-effort (CPUE) declines with increased effort due to the stoc externality. Catches from tuna ( C i, 1 ) and swordfish sets ( C i, 2 ) in area can be separated from (A4) as C = B α γ E (A6) i, 1 i, i, i1 1 C = B α γ E, (A7) i, 2 i, i, i2 2 25

26 where C + C = C. Dividing (A7) by (A6), we obtain i, 1 i, 2 i, γ i γ 2 i1 = CPUE CPUE i, 2 i, 1 (A8) where CPUE = C E is CPUE for species i in area for the tuna set (similarly for the i, 1 i, 1 i, 1 swordfish set). In (A8), both catchability coefficients (γ i1 and γ i2 ) are assumed constant, while CPUE is expected to fluctuate over seasons and vary across fishing locations. We can now estimate the catchability ratio (γ i2 /γ i1 ) from the CPUE ratio in (A8). Since the noise associated with CPUE data is expected to be large due to fluctuations in stoc size within each month and non-uniformity of fish stocs within each location (five-by-five degree square), we only use data points that consist of at least 30 sets each of the tuna and swordfish sets. The resulting estimates of the catchability ratio for the five species are presented in Table 1. Oamoto (1999) computed the average CPUE ratios by swordfish and tuna sets and obtained similar results. Iterative Procedure of OLS Estimation for Catchability Coefficients Suppose only one type of fishing strategy is used to catch a single species, then given the catch function (A4), a series of harvests in a fishing location over consecutive periods yields the following sequence of catches: C = γ α E B Μ Μ C = γ α E B (A9) t t t t ( ) C = γ α E B = γ α E B C + R t + 1 t + 1 t + 1 t + 1 t + 1 t + 1 t t t where all subscripts denote the time period, and subscripts denoting species, area, and fishing strategy are omitted for notational simplicity. The net stoc inflow during period t, R t, is defined as the total fish stoc inflow minus the stoc outflow and natural mortality. It is positive if the inflow is greater than the outflow, and vice versa. Note that in (A9), the fish stoc changes 26